The U.S. government has significantly underestimated the number of jobs created this year as it struggled to analyze data distorted by the effects of the pandemic, creating new challenges for policy makers navigating a very volatile economic environment.
During 2021, the government agency that publishes monthly US employment statistics revised its initial estimates of wage growth upwards by a total of 976,000 jobs, the largest such adjustment in a single year.
Reviewing past estimates to take into account new data is a routine task: the first estimate calculated by the Bureau of Labor Statistics (BLS) is updated twice, with the second estimate in the next month’s job report and the third in the following month is published.
However, at key economic tipping points, these numbers are not only the most difficult to measure, but they are also the most closely monitored. This means that data revisions, which are normally published without fanfare, suddenly have the potential to fundamentally change the perception of the health of the labor market and with it the precise fine-tuning of politics.
“Economic fundamentals have changed at an unprecedented rate. We saw not in my lifetime and not in the lifetimes of most of the people living today. . . an economic recovery that has been so rapid since spring 2020, “said David Wilcox, former head of research and statistics at the Federal Reserve. “The challenges of economic measurement in a pandemic environment are enormous.”
The pandemic has proven uniquely delicate for economists looking at the US labor market, as it has not only resulted in delays in data reporting, but has also skewed the seasonal rhythms that are normally included in estimates. The sheer magnitude of wage growth during the economic recovery has also created challenges.
“You are taking an already difficult task and making it absolutely herculean,” said Ernie Tedeschi, a senior economic advisor to the Biden administration as a member of the economic council.
This difficulty was shown once again in the latest job report from December 3, which only showed 210,000 new jobs created in November. The unemployment rate fell further to 4.2 percent, far from almost 15 percent in the previous year.
The numbers baffled economists, not least because of a sizeable gap between the two surveys that make up the monthly report. The company survey “Companies”, from which the revised headline numbers are derived, indicated a clear slowdown in new hires. The “Household” personal survey, which was used to calculate the unemployment rate, showed an increase of 1.1 million euros.
The picture got even more complicated when the BLS comprehensively updated its September and October numbers, bringing the job growth to 379,000 in September and 546,000 in October. Taken together, around 6 million jobs were regained in 2021.
“Everything was so much bigger than it was before,” says Stephen Crestol, an economist who has spent more than three decades at BLS. “We are not used to a big decline and big increases.”
The task of estimating wages during the pandemic was made difficult by two main factors. First, more and more companies are giving their survey responses too late.
The agency struggled to get companies to choose the charter survey during the pandemic, Crestol said. Participation, which is voluntary in all but three states, has halved since last February.
Even if companies agree to participate in the survey, their responses can be received in the job report, which is published on the first Friday of the month, after the deadline for the first estimate has passed.
For the November report, 65.3 percent of companies replied in time, the lowest value for a November report in more than a decade.
To create the first number, the BLS calculates an expected salary number for companies that have not yet responded. Late responses will be reflected in subsequent revisions.
“The bottom line is that you don’t get that much change in a normal month,” said Cornell University’s Erica Groshen and former BLS commissioner. But in a year like 2021, which is marked by an extremely rapid economic recovery, the slowest reacting companies could be the quickest to hire, resulting in major upward revisions.
Seasonal adjustment factors have also made measuring employment growth a tedious task.
Economists take a close look at the “seasonally adjusted” figures because they are supposed to offer the most direct insight into current developments when regular fluctuations in data associated with events such as the start of the school year are factored out.
Retail companies, for example, had 331,600 more people on their payroll in November than in October. However, the BLS has revised that number down as retail attitudes tend to pick up just before the holiday season begins. The agency reported that the industry “lost” 20,400 jobs, seasonally adjusted.
Overall, the raw numbers in November showed a wage increase of 778,000, which the BLS revised down by 568,000, a record revision.
“The seasonal patterns are not law, just patterns,” said Betsey Stevenson, a professor of economics at the University of Michigan and a member of the CEA during the Obama administration. “Covid did more to break our patterns than I could have imagined.”
Seasonal factors are based on data for the past 10 years, with more emphasis being placed on recent years. In 2020, many movements were considered outliers and were not included in the adjustment model.
The model is constantly optimized as new data comes in, which leads to further revisions. “As you get more data, it will affect your understanding of past seasonal trends,” said Nick Bunker of Indeed, the job website.
How the new coronavirus variant develops will determine whether the latest waves of revisions increase or become larger from here.
The impact is huge for the Federal Reserve, which is closely monitoring the employment situation to get the green light to tighten monetary policy next year.
Christopher Waller, a Fed governor, said the central bank had already supplemented its models with radio frequency data and other sources, including the weekly ADP employment report.
“The more we see that we are recovering from a recession, the better the statistical methods will work,” said Groshen. “All the more so since it is being driven by a new variant and a different political response to it. . . the more the models can fail because they are based on the past. “